# alembic-postgresql-enum
[<img src="https://img.shields.io/pypi/pyversions/alembic-postgresql-enum">](https://pypi.org/project/alembic-postgresql-enum/)
[<img src="https://img.shields.io/pypi/v/alembic-postgresql-enum">](https://pypi.org/project/alembic-postgresql-enum/)
[<img src="https://img.shields.io/pypi/l/alembic-postgresql-enum">](https://pypi.org/project/alembic-postgresql-enum/)

Alembic autogenerate support for creation, alteration and deletion of enums

Alembic will now automatically:
- Create enums that currently are not in postgres schema
- Remove/add/alter enum values
- Reorder enum values
- Delete unused enums from schema

If you are curious to know about analogs and reasons for this library to exist see [alternatives and motivation](https://github.com/Pogchamp-company/alembic-postgresql-enum/blob/master/docs/alternatives.md)

## Usage

Install library:
```
pip install alembic-postgresql-enum
```

Add the line:

```python 
# env.py
import alembic_postgresql_enum
...
```

To the top of your migrations/env.py file.

## Features

* [Creation of enums](#creation-of-enum)
* [Deletion of unreferenced enums](#deletion-of-unreferenced-enum)
* [Detection of enum values changes](#detection-of-enum-values-changes)
  * [Creation of new enum values](#creation-of-new-enum-values)
  * [Deletion of enums values](#deletion-of-enums-values)
  * [Renaming of enum values](#rename-enum-value)
* [Omitting managing enums](#omitting-managing-enums)

## Creation of enum<a id="creation-of-enum"></a>

### When table is created

```python
class MyEnum(enum.Enum):
    one = 1
    two = 2
    three = 3


class ExampleTable(BaseModel):
    test_field = Column(Integer, primary_key=True, autoincrement=False)
    enum_field = Column(postgresql.ENUM(MyEnum)) 
```
This code will generate migration given below: 
```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    # this line is generated by our library
    sa.Enum('one', 'two', 'three', name='myenum').create(op.get_bind())
    op.create_table('example_table',
    sa.Column('test_field', sa.Integer(), nullable=False),
    # create_type=False argument is now present on postgresql.ENUM as library takes care of enum creation
    sa.Column('enum_field', postgresql.ENUM('one', 'two', 'three', name='myenum', create_type=False), nullable=True),
    sa.PrimaryKeyConstraint('test_field')
    )
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    # drop_table does not drop enum by alembic
    op.drop_table('example_table')
    # It is dropped by us
    sa.Enum('one', 'two', 'three', name='myenum').drop(op.get_bind())
    # ### end Alembic commands ###
```

### When column is added
```python
class MyEnum(enum.Enum):
    one = 1
    two = 2
    three = 3


class ExampleTable(BaseModel):
    test_field = Column(Integer, primary_key=True, autoincrement=False)
    # this column has just been added
    enum_field = Column(postgresql.ENUM(MyEnum)) 
```
This code will generate migration given below: 
```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    # this line is generated by our library
    sa.Enum('one', 'two', 'three', name='myenum').create(op.get_bind())
    # create_type=False argument is now present on postgresql.ENUM as library takes care of enum creation
    op.add_column('example_table', sa.Column('enum_field', postgresql.ENUM('one', 'two', 'three', name='myenum', create_type=False), nullable=False))
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.drop_column('example_table', 'enum_field')
    # enum is explicitly dropped as it is no longer used
    sa.Enum('one', 'two', 'three', name='myenum').drop(op.get_bind())
    # ### end Alembic commands ###
```

## Deletion of unreferenced enum<a id="deletion-of-unreferenced-enum"></a>
If enum is defined in postgres schema, but its mentions removed from code - It will be automatically removed
```python
class ExampleTable(BaseModel):
    test_field = Column(Integer, primary_key=True, autoincrement=False)
    # enum_field is removed from table
```

```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.drop_column('example_table', 'enum_field')
    sa.Enum('one', 'two', 'four', name='myenum').drop(op.get_bind())
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    sa.Enum('one', 'two', 'four', name='myenum').create(op.get_bind())
    op.add_column('example_table', sa.Column('enum_field', postgresql.ENUM('one', 'two', 'four', name='myenum', create_type=False), autoincrement=False, nullable=True))
    # ### end Alembic commands ###
```

## Detection of enum values changes<a id="detection-of-enum-values-changes"></a>

***Can be disabled with `detect_enum_values_changes` configuration flag turned off***

### Creation of new enum values<a id="creation-of-new-enum-values"></a>

If new enum value is defined sync_enum_values function call will be added to migration to account for it

```python
class MyEnum(enum.Enum):
    one = 1
    two = 2
    three = 3
    four = 4 # New enum value
```

```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'three', 'four'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'three'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###
```

### Deletion of enums values<a id="deletion-of-enums-values"></a>

If enum value is removed it also will be detected

```python
class MyEnum(enum.Enum):
    one = 1
    two = 2
    # three = 3 removed
```

```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'three'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###
```


### Rename enum value<a id="rename-enum-value"></a>
In this case you must manually edit migration

```python
class MyEnum(enum.Enum):
    one = 1
    two = 2
    three = 3 # renamed from `tree`
```

This code will generate this migration:
```python
def upgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'three'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###


def downgrade():
    # ### commands auto generated by Alembic - please adjust! ###
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'tree'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[],
    )
    # ### end Alembic commands ###
```

This migration will cause problems with existing rows that references MyEnum

So adjust migration like that

```python
def upgrade():
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'three'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[('tree', 'three')],
    )


def downgrade():
    op.sync_enum_values(
        enum_schema='public', 
        enum_name='myenum', 
        new_values=['one', 'two', 'tree'], 
        affected_columns=[TableReference(table_schema='public', table_name='example_table', column_name='enum_field')],
        enum_values_to_rename=[('three', 'tree')],
    )
```

Do not forget to switch places old and new values for downgrade

All defaults in postgres will be renamed automatically as well

## Omitting managing enums<a id="omitting-managing-enums"></a>

If configured `include_name` function returns `False` given enum will be not managed.
```python
import alembic_postgresql_enum

def include_name(name: str) -> bool:
    return name not in ['enum-to-ignore', 'some-internal-enum']

alembic_postgresql_enum.set_configuration(
    alembic_postgresql_enum.Config(
        include_name=include_name,
    )
)
```

Feature is similar to [sqlalchemy feature for tables](https://alembic.sqlalchemy.org/en/latest/autogenerate.html#omitting-table-names-from-the-autogenerate-process)

## Configuration

You can configure this extension to disable parts of it, or to enable some feature flags

To do so you need to call set_configuration function after the import:

```python
import alembic_postgresql_enum

alembic_postgresql_enum.set_configuration(
    alembic_postgresql_enum.Config(
        add_type_ignore=True,
    )
)
```

Available options:

- `add_type_ignore` (`False` by default) - flag that can be turned on 
to add `# type: ignore[attr-defined]` at the end of generated `op.sync_enum_values` calls.
This is helpful if you are using type checker such as `mypy`.
`type: ignore` is needed because there is no way to add new function to an existing alembic's `op`.

- `include_name` (`lambda _: True` bby default) - it adds ability to ignore process enum by name in similar way alembic allows to define `include_name` function. 
This property accepts function that takes enum name and returns whether it should be processed.  

- `drop_unused_enums` (`True` by default) - feature flag that can be turned off to disable clean up of undeclared enums

- `detect_enum_values_changes` (`True` by default) - feature flag that can be turned off to disable generation of `op.sync_enum_values`.